India – Pakistan conflict

One 0–100 scale for the whole conflict. 0–49 is escalation before war; 50–100 is active war and how far it has gone.

50 is not a threshold the number can cross on its own. Which side a theater sits on is decided by whether verified belligerents are fighting there, so war onset always moves the score up across 50 — the number never declares a war by climbing.

It is a stage, not a probability: 42 is not a 42% chance of war. Below 50 it grades movement toward war, above 50 the war itself.

Recomputed every two days while a theater is at war and every five otherwise, from open reporting by fixed formula — the model never picks a number. Between recomputations the index holds its last value rather than being re-judged without new evidence. Full construction →

Direction — which way the week pushed. Severity — where the index sits. Connections — how much of the rest a shock here reaches. Cascade — that reach combined with how ready this theater is to go off.

The index reads the state of the conflict, not the volume of news. Each run grades the factors listed below against fixed rubrics, and the score follows by formula.

So a dramatic headline moves nothing unless it changes a grade. A theater already graded for recurring incursions absorbs one more without moving — it is priced in.

Open How this was measured for every grade, its movement since the last run, and the events behind it. Full construction →

A different measurement from the score above, from a different source. This line counts how much of the world's news reports conflict between these two states, against their own long-run normal: 100 is normal, 200 is twice the usual level.

It is computed mechanically from the GDELT Project's open event stream every 15 minutes and never touches the assessed index above — when both move together, that agreement is the signal. A day with too few events shows no reading rather than a guess. Click the chart for the full history.

Data: the GDELT Project.

Possible new links between this theater and the others, found automatically in machine-coded news: a state acting in both, alliance bridges, retaliation across them. Each entry leads with the headline that evidenced it.

Discovery, not verification. Nothing here changes any score. Nothing appears unless at least two independent sources from two different domains reported it within 72 hours, and only the editorial process can mark a connection verified.